Computational Prediction of Human Disease- Associated circRNAs Based on Manifold Regularization Learning Framework.

The accumulating evidences regarding circular RNAs (circRNAs) indicate that they play crucial roles in a wide range of biological processes and participate in tumorigenesis and progression. The number of newly discovered circRNAs have increased dramatically in recent years, but the functions of vast majority of circRNAs remain unknown, and little effort has been devoted to discover disease-associated circRNAs on a large scale until now. With the advancement of high-throughput technology, the increasing availability of omics data has provided an unprecedented opportunity for prioritizing candidate circRNAs for diseases by computational models, which will contribute to exploring the pathogenesis of complex diseases at the circRNA level and provide promising applications in disease diagnosis and treatment. Here we propose the assumption that circRNAs with similar functions are normally associated with similar diseases and vice versa, and develop an integrated computational framework called MRLDC to identify disease-associated circRNAs. To our knowledge, little efforts have been developed for uncovering circRNA-disease associations on a large scale. By fully exploiting the experimentally validated associations between diseases and circRNAs, we first compute the Gaussian interaction profile kernel similarity for circRNAs and diseases, and then a heterogeneous circRNA-disease bilayer network is constructed by combining a circRNA similar network, a disease similar network, and known circRNA-disease associations. Subsequently, we develop a weighted low-rank approximation optimization algorithm with dual-manifold regularizations for predicting disease-associated circRNAs. Experimental results indicate that MRLDC can effectively identify disease circRNA candidates with high accuracy. In addition, case studies further demonstrate the ability of our method in discovering potential circRNA-disease associations.